Strength or Accuracy? Fitness Calculation in Learning Classifier Systems
Learning Classifier Systems, From Foundations to Applications
What Makes a Problem Hard for XCS?
IWLCS '00 Revised Papers from the Third International Workshop on Advances in Learning Classifier Systems
Genetic Programming and Evolvable Machines
Performance and population state metrics for rule-based learning systems
CEC '02 Proceedings of the Evolutionary Computation on 2002. CEC '02. Proceedings of the 2002 Congress - Volume 02
Classifier fitness based on accuracy
Evolutionary Computation
Functional Verification Coverage Measurement and Analysis
Functional Verification Coverage Measurement and Analysis
Learning classifier systems: a complete introduction, review, and roadmap
Journal of Artificial Evolution and Applications
Analysis of the niche genetic algorithm in learning classifier systems
Proceedings of the 15th annual conference on Genetic and evolutionary computation
Comparison of two methods for computing action values in XCS with code-fragment actions
Proceedings of the 15th annual conference companion on Genetic and evolutionary computation
| Hi-index | 0.00 |
In this paper we applied the eXtended Classifier System (XCS) on a novel real world problem, namely digital Design Verification (DV). We witnessed the inadequacy of XCS on binary problems that contain high overlap between optimal rules especially when the focus is on population and not system level performance. The literature attempts to underplay the importance of the aforementioned weakness and in short, supports that a) XCS can potentially learn any Boolean function given enough resources are allocated (right parameters used) and b) the main metric deciding the learning difficulty of a Boolean function is the amount of classifiers required to represent it (i.e. |[O]|). With this work we experimentally refuted the aforementioned propositions and as a result of the work, we introduce new insights on the behavior of XCS when solving two-valued Boolean functions using a binary reward scheme (1000/0). We also introduce a new population metric (%[EPI]) that should necessarily be used to guide future research on improving XCS performance on the aforementioned problems.